Case studies

Real problems. Working solutions.

A few examples of the kinds of systems we build. Names and identifying details are changed to protect the organizations involved. What matters is the pattern, and how it might apply to you.

Private AI · Microsoft Teams

Inbox to Action

A private AI assistant that turns everyday chat requests into tracked, owner-assigned tasks.

0new tools staff had to learn
100%on-network: no data sent to outside AI
Local LLMMicrosoft TeamsPythonSQL Server

The challenge

At a government-services firm, requests arrived the way they always had: casually, in Teams chats. Things slipped. Previous attempts to fix it meant asking everyone to use a ticketing system, and nobody did.

What we built

  • A language model hosted on the company’s own server reads incoming Teams messages; nothing leaves the network.
  • A two-stage pipeline first decides whether a message is actually a request, then extracts the task, owner and due date as structured data.
  • When details are missing, it asks the requester right in Teams and matches their reply back to the pending task.
  • An approval workflow and scheduled email summaries keep managers in the loop.

Why it stuck

Nobody had to change how they ask for things. The system adapted to the team, not the other way around.

Automated reporting & alerts

Contract Pulse

A finance early-warning platform that replaced weeks-late manual reports with live data and automatic alerts.

75contracts monitored automatically
13alert types delivered by email
ETL pipelineSQL ServerMicrosoft GraphDashboards

The challenge

Leadership tracked spending across dozens of government contracts with manually built reports. By the time an overrun showed up, it was weeks old and expensive.

What we built

  • An automated pipeline pulls billed-labor data from the company’s ERP twice a week, about 38,600 records each run, with no one exporting anything.
  • A database engine calculates burn rate, estimate-at-completion and staffing for every contract.
  • Change detection sends 13 kinds of budget and schedule alerts straight to the right inbox.
  • Dashboards are there for anyone who wants to dig in, but no one has to go looking.

Why it stuck

The alerts come to people in Outlook, where they already work. Nobody had to remember to check a new system.

Forecasting & analytics

Capacity Lens

Schedule and staffing forecasting that found the one bottleneck behind a decades-long overrun.

99.7%match to the official scheduling engine
~2sto re-plan the entire program
PythonMonte CarloForecastingWeb app

The challenge

An aerospace manufacturing program ran on a schedule of thousands of tasks, with staffing estimated in a hand-maintained spreadsheet. The forecast didn’t add up, and nobody could say why.

What we built

  • A critical-path engine that reproduces the scheduling software’s own date math on 99.7% of tasks.
  • Monte Carlo risk modeling showing the confidence behind every committed date.
  • A resource-constrained forecasting app that re-plans about 4,900 tasks in roughly two seconds.
  • An automated staffing pipeline that replaced the spreadsheet and exposed a hidden modeling error.

Why it stuck

It answered the question leadership actually had: the delay traced back to a single understaffed trade, with a sized fix.

Automation rescue

Legacy Tool Rescue

A business-critical desktop add-in, inherited from a previous team, stabilized and extended.

58legacy modules brought under control
13+production defects closed
VBAVB.NETGitRoot-cause analysis

The challenge

A company’s flagship productivity add-in had been built by a team that had since moved on. It crashed, had no version control, and nobody fully understood it.

What we built

  • Migrated 58 legacy modules into a version-controlled, modular codebase with centralized error handling.
  • Closed 13+ production defects through a structured testing workflow.
  • Traced the worst crash through crash-dump analysis to an undocumented defect in the host software itself.
  • Added a new interactive visualization feature on top of the now-stable foundation.

Why it stuck

Existing users kept their familiar tool. It simply stopped breaking, then started doing more.

Microsoft 365 & Power Platform

Everyday Automations

Small, targeted automations that quietly remove friction from dozens of people’s days.

Daysnot months, to deliver each one
M365built on tools already licensed
Power AppsPower AutomateSharePointPower BI

The challenge

Not every win is a big platform. Most organizations lose hours to small chores: expense receipts, timesheets, verification codes, status emails.

What we built

  • A receipt-scanner app that uses AI to read receipts and log them automatically.
  • An hours logger with automatic reminders that improved timesheet accuracy.
  • Two-factor verification codes routed into Teams so a shared login stops being a bottleneck.
  • Alerts when documents move between project phases, and dynamic recurring reminders.

Why it stuck

Each one takes a few minutes off someone’s day. Across a whole organization, that adds up to real capacity.